Trump Signs Revised AI Executive Order with 30-Day Gov Access to Frontier Models

President Donald Trump signed a scaled-back AI executive order on Monday night, granting the US government exclusive 30-day access to the most advanced AI models before public release. The revised order, which replaced an earlier 90-day proposal shelved on May 21, targets frontier systems like Anthropic's Claude Mythos and OpenAI's GPT-5.5. The voluntary framework, driven by White House Chief of Staff Susie Wiles, Treasury Secretary Scott Bessent, and National Cyber Director Sean Cairncross, bypasses former AI czar David Sacks' skepticism of government intervention. The Pentagon must shore up classified networks within 30 days, and the Justice Department will prosecute AI-enabled hacking. AI companies including Anthropic expressed support, while Bessent plans to explore a similar cross-border framework with China.

The order emerged after intense internal debates: executives from top AI labs convinced the White House that delaying regulation was untenable as models grew more powerful. Trump approved the 30-day window only after securing enough industry buy-in. Federal agencies will now create a classified process to determine which models require early government access, with select 'trusted partners' also granted previews. The directive stops short of formal regulation but marks the first major AI policy move of Trump's second term. White House spokesperson Liz Huston framed it as a 'common-sense approach collaborating with industry to balance innovation and security.' Critics note the shortened review period may still burden rapidly evolving AI development cycles.